{
  "uid": "atlas-nodes",
  "title": "Atlas Nodes",
  "folderUid": "atlas-internal",
  "editable": true,
  "panels": [
    {
      "id": 1,
      "type": "stat",
      "title": "Worker Nodes Ready",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 4,
        "w": 8,
        "x": 0,
        "y": 0
      },
      "targets": [
        {
          "expr": "sum(kube_node_status_condition{condition=\"Ready\",status=\"true\",node=~\"titan-04|titan-05|titan-06|titan-07|titan-08|titan-09|titan-10|titan-11|titan-12|titan-13|titan-14|titan-15|titan-16|titan-17|titan-18|titan-19|titan-22|titan-24\"})",
          "refId": "A"
        }
      ],
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "rgba(115, 115, 115, 1)",
                "value": null
              },
              {
                "color": "green",
                "value": 1
              }
            ]
          },
          "unit": "none",
          "custom": {
            "displayMode": "auto",
            "valueSuffix": "/18"
          }
        },
        "overrides": []
      },
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "center",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "textMode": "value"
      }
    },
    {
      "id": 2,
      "type": "stat",
      "title": "Control Plane Ready",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 4,
        "w": 8,
        "x": 8,
        "y": 0
      },
      "targets": [
        {
          "expr": "sum(kube_node_status_condition{condition=\"Ready\",status=\"true\",node=~\"titan-0a|titan-0b|titan-0c\"})",
          "refId": "A"
        }
      ],
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "rgba(115, 115, 115, 1)",
                "value": null
              },
              {
                "color": "green",
                "value": 1
              }
            ]
          },
          "unit": "none",
          "custom": {
            "displayMode": "auto",
            "valueSuffix": "/3"
          }
        },
        "overrides": []
      },
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "center",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "textMode": "value"
      }
    },
    {
      "id": 3,
      "type": "stat",
      "title": "Control Plane Workloads",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 4,
        "w": 8,
        "x": 16,
        "y": 0
      },
      "targets": [
        {
          "expr": "sum(kube_pod_info{node=~\"titan-0a|titan-0b|titan-0c\",namespace!~\"kube-system|kube-public|kube-node-lease|longhorn-system|monitoring|flux-system\"})",
          "refId": "A"
        }
      ],
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "rgba(115, 115, 115, 1)",
                "value": null
              },
              {
                "color": "green",
                "value": 1
              }
            ]
          },
          "unit": "none",
          "custom": {
            "displayMode": "auto"
          }
        },
        "overrides": []
      },
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "center",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "textMode": "value"
      }
    },
    {
      "id": 9,
      "type": "stat",
      "title": "API Server 5xx rate",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 4,
        "w": 8,
        "x": 0,
        "y": 4
      },
      "targets": [
        {
          "expr": "sum(rate(apiserver_request_total{code=~\"5..\"}[5m]))",
          "refId": "A"
        }
      ],
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "yellow",
                "value": 0.05
              },
              {
                "color": "orange",
                "value": 0.2
              },
              {
                "color": "red",
                "value": 0.5
              }
            ]
          },
          "unit": "req/s",
          "custom": {
            "displayMode": "auto"
          },
          "decimals": 3
        },
        "overrides": []
      },
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "center",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "textMode": "value"
      }
    },
    {
      "id": 10,
      "type": "stat",
      "title": "API Server P99 latency",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 4,
        "w": 8,
        "x": 8,
        "y": 4
      },
      "targets": [
        {
          "expr": "histogram_quantile(0.99, sum by (le) (rate(apiserver_request_duration_seconds_bucket[5m]))) * 1000",
          "refId": "A"
        }
      ],
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "yellow",
                "value": 250
              },
              {
                "color": "orange",
                "value": 400
              },
              {
                "color": "red",
                "value": 600
              }
            ]
          },
          "unit": "ms",
          "custom": {
            "displayMode": "auto"
          },
          "decimals": 1
        },
        "overrides": []
      },
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "center",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "textMode": "value"
      }
    },
    {
      "id": 11,
      "type": "stat",
      "title": "etcd P99 latency",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 4,
        "w": 8,
        "x": 16,
        "y": 4
      },
      "targets": [
        {
          "expr": "histogram_quantile(0.99, sum by (le) (rate(etcd_request_duration_seconds_bucket[5m]))) * 1000",
          "refId": "A"
        }
      ],
      "fieldConfig": {
        "defaults": {
          "color": {
            "mode": "thresholds"
          },
          "mappings": [],
          "thresholds": {
            "mode": "absolute",
            "steps": [
              {
                "color": "green",
                "value": null
              },
              {
                "color": "yellow",
                "value": 50
              },
              {
                "color": "orange",
                "value": 100
              },
              {
                "color": "red",
                "value": 200
              }
            ]
          },
          "unit": "ms",
          "custom": {
            "displayMode": "auto"
          },
          "decimals": 1
        },
        "overrides": []
      },
      "options": {
        "colorMode": "value",
        "graphMode": "area",
        "justifyMode": "center",
        "reduceOptions": {
          "calcs": [
            "lastNotNull"
          ],
          "fields": "",
          "values": false
        },
        "textMode": "value"
      }
    },
    {
      "id": 4,
      "type": "timeseries",
      "title": "Node CPU",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 9,
        "w": 24,
        "x": 0,
        "y": 8
      },
      "targets": [
        {
          "expr": "avg by (node) (((1 - avg by (instance) (rate(node_cpu_seconds_total{mode=\"idle\"}[5m]))) * 100) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))",
          "refId": "A",
          "legendFormat": "{{node}}"
        }
      ],
      "fieldConfig": {
        "defaults": {
          "unit": "percent"
        },
        "overrides": []
      },
      "options": {
        "legend": {
          "displayMode": "table",
          "placement": "right",
          "calcs": [
            "last"
          ]
        },
        "tooltip": {
          "mode": "multi"
        }
      }
    },
    {
      "id": 5,
      "type": "timeseries",
      "title": "Node RAM",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 9,
        "w": 24,
        "x": 0,
        "y": 17
      },
      "targets": [
        {
          "expr": "avg by (node) ((avg by (instance) ((node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes * 100)) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))",
          "refId": "A",
          "legendFormat": "{{node}}"
        }
      ],
      "fieldConfig": {
        "defaults": {
          "unit": "percent"
        },
        "overrides": []
      },
      "options": {
        "legend": {
          "displayMode": "table",
          "placement": "right",
          "calcs": [
            "last"
          ]
        },
        "tooltip": {
          "mode": "multi"
        }
      }
    },
    {
      "id": 6,
      "type": "timeseries",
      "title": "Control Plane (incl. titan-db) CPU",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 9,
        "w": 12,
        "x": 0,
        "y": 26
      },
      "targets": [
        {
          "expr": "(avg by (node) (((1 - avg by (instance) (rate(node_cpu_seconds_total{mode=\"idle\"}[5m]))) * 100) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))) * on(node) group_left() label_replace(node_uname_info{nodename=~\"titan-0a|titan-0b|titan-0c|titan-db\"}, \"node\", \"$1\", \"nodename\", \"(.*)\")",
          "refId": "A",
          "legendFormat": "{{node}}"
        }
      ],
      "fieldConfig": {
        "defaults": {
          "unit": "percent"
        },
        "overrides": []
      },
      "options": {
        "legend": {
          "displayMode": "table",
          "placement": "right"
        },
        "tooltip": {
          "mode": "multi"
        }
      }
    },
    {
      "id": 7,
      "type": "timeseries",
      "title": "Control Plane (incl. titan-db) RAM",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 9,
        "w": 12,
        "x": 12,
        "y": 26
      },
      "targets": [
        {
          "expr": "(avg by (node) ((avg by (instance) ((node_memory_MemTotal_bytes - node_memory_MemAvailable_bytes) / node_memory_MemTotal_bytes * 100)) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))) * on(node) group_left() label_replace(node_uname_info{nodename=~\"titan-0a|titan-0b|titan-0c|titan-db\"}, \"node\", \"$1\", \"nodename\", \"(.*)\")",
          "refId": "A",
          "legendFormat": "{{node}}"
        }
      ],
      "fieldConfig": {
        "defaults": {
          "unit": "percent"
        },
        "overrides": []
      },
      "options": {
        "legend": {
          "displayMode": "table",
          "placement": "right"
        },
        "tooltip": {
          "mode": "multi"
        }
      }
    },
    {
      "id": 8,
      "type": "timeseries",
      "title": "Root Filesystem Usage",
      "datasource": {
        "type": "prometheus",
        "uid": "atlas-vm"
      },
      "gridPos": {
        "h": 9,
        "w": 24,
        "x": 0,
        "y": 35
      },
      "targets": [
        {
          "expr": "avg by (node) ((avg by (instance) ((1 - (node_filesystem_avail_bytes{mountpoint=\"/\",fstype!~\"tmpfs|overlay\"} / node_filesystem_size_bytes{mountpoint=\"/\",fstype!~\"tmpfs|overlay\"})) * 100)) * on(instance) group_left(node) label_replace(node_uname_info{nodename!=\"\"}, \"node\", \"$1\", \"nodename\", \"(.*)\"))",
          "refId": "A",
          "legendFormat": "{{node}}"
        }
      ],
      "fieldConfig": {
        "defaults": {
          "unit": "percent"
        },
        "overrides": []
      },
      "options": {
        "legend": {
          "displayMode": "table",
          "placement": "right"
        },
        "tooltip": {
          "mode": "multi"
        }
      },
      "timeFrom": "30d"
    }
  ],
  "time": {
    "from": "now-12h",
    "to": "now"
  },
  "annotations": {
    "list": []
  },
  "schemaVersion": 39,
  "style": "dark",
  "tags": [
    "atlas",
    "nodes"
  ]
}